OpenAI’s ambitious 28-day shipping sprint for its Codex platform has yielded its first official milestone, though subscribers hoping for an immediate, complimentary reset of their usage quotas must continue to exercise patience. Thibault Sottiaux, the engineering lead for Codex at OpenAI, announced the debut update on the opening day of the challenge, confirming that foundational speed improvements have been successfully deployed across the company’s ecosystem.
The initial rollout focuses primarily on performance optimization. Sottiaux reported that the GPT-6 Astra and GPT-6.1 Sol models are now operating approximately 50 percent faster by default. This enhancement marks the inaugural delivery in an unusually demanding challenge Sottiaux initiated over the weekend, wherein he pledged to ship a verifiable, clear improvement relevant to the majority of Codex and ChatGPT Work users every single day for 28 consecutive days. Alternatively, the engineering lead committed to executing a full, global usage reset if the team fails to meet its daily deployment goals.
Accelerated Performance for GPT-6 Astra and Sol Models
Elaborating on the technical adjustments in a subsequent statement, Sottiaux quantified the performance gains, noting that both the Astra and Sol models are now capable of generating approximately 50 tokens per second, a significant increase from the previous baseline of roughly 30 tokens per second. While this output represents a jump of closer to two-thirds rather than a strict 50 percent improvement, both figures serve as early field estimates ahead of formal benchmark evaluations.
Crucially, these efficiency gains are not restricted to OpenAI’s proprietary interfaces. Subscribers authenticating through Sign in with ChatGPT are also experiencing the accelerated speeds via third-party integrations, including OpenCode, Pi, Amp, and Devin.
The timing of this performance update introduces a notable element of irony within the development team. Just 48 hours prior to the commencement of the 28-day challenge, Sottiaux issued a public apology regarding a sluggish rollout for GPT-6.1 Sol. He attributed the initial latency to a substantial, unexpected load spike during the model’s first two days on the market. To compensate for the disruption, OpenAI announced a global usage reset for all paid ChatGPT accounts. Consequently, the very first deliverable of the new daily sprint is a direct performance enhancement applied to the model that caused the initial service bottleneck.
Furthermore, this initial deployment establishes a working definition of what OpenAI leadership considers a qualifying ship for the challenge. Just before kicking off the sprint, Sottiaux outlined the strategic pillars guiding the next four weeks, emphasizing a dedicated focus on simplification, efficiency translating into higher throughput, groundbreaking features, and new model architectures. He underscored that user feedback consistently points toward a strong industry preference for simplified workflows and reduced friction.
The Strategic Importance of Usage Resets as Currency
Within the Codex ecosystem, usage rate limits function effectively as a form of digital currency. Over the past several months, OpenAI has conditioned its subscriber base to monitor and anticipate usage resets during service disruptions or milestones.
In April, when Sottiaux reset Codex rate limits across all paid tiers, he humorously noted in his announcement that resets are costly and should not be distributed casually, before ultimately conceding to positive user sentiment and community demand. The frequency of these resets increased over the summer. In July, following the integration of Codex into the standard ChatGPT desktop app—a transition that inadvertently introduced regressions in specific multi-agent development workflows—OpenAI executed two separate usage resets within a single day.
Subsequent milestones, such as Codex and ChatGPT Work surpassing eight million active users, were similarly commemorated with quota renewals. The practice culminated in a live, on-stage reset handed out by Sottiaux and OpenAI CEO Sam Altman during the company’s DevDay keynote event on September 29.
This recurring pattern has fostered an active sub-ecosystem of third-party trackers. Within hours of Sottiaux’s announcement of the 28-day challenge, independent developers launched a dedicated 28-day scoreboard to monitor whether the engineering team successfully ships a feature or defaults to a reset.
The community reaction has been a mixture of anticipation and skepticism. Some developers have jokingly compared the daily sprint to a high-stakes gambling mechanism, encouraging subscribers to intentionally exhaust their daily allowances in the hope that the engineering team misses a shipping target. Conversely, professional users managing tight production schedules have expressed a preference for higher standard limits rather than sporadic resets. A representative Plus subscriber noted that the existing five-hour usage cap frequently dictates and restricts professional engineering workflows, while software engineers at enterprise firms have treated the pledge as a guaranteed pathway to 28 consecutive days of refreshed quotas.
Enterprise Pricing Adjustments and Competitive Pressures
The ongoing discourse surrounding usage allowances coincides with structural changes to OpenAI’s enterprise and professional subscription tiers. OpenAI announced that near the conclusion of the 28-day window—specifically beginning October 30—the company will adjust the Codex and ChatGPT Work allowance on its $200 Pro plan. The quota will be reduced from 20 times the standard Plus allowance down to 10 times.
To accommodate heavy users affected by this reduction, OpenAI introduced a new $500 enterprise tier offering 25 times the Plus allowance, alongside a specialized Ultrafast tier for GPT-6 Astra. The Ultrafast tier promises up to eight times the standard processing speed within Codex environments, creating a clear stratification between casual professional use and intensive enterprise development.
Simultaneously, OpenAI faces mounting competitive pressures within the AI-assisted coding market. While Sottiaux’s daily sprint aims to maintain user engagement, critics and community members argue that incremental updates may prove insufficient against formidable rivals. Pseudonymous industry analysts have suggested that 28 minor operational improvements will carry limited weight unless OpenAI delivers a flagship model capable of decisively outperforming Anthropic’s Claude Opus 5.5.
Market adoption metrics underscore this competitive challenge. Prominent developer ecosystem figures reported that Anthropic’s newest Opus model has rapidly captured significant market share, crossing a threshold where nearly half of all programming prompts within specific development environments are directed toward Anthropic’s technology rather than OpenAI’s offerings.
Chronology, Governance, and Accountability of the Challenge
A notable characteristic of the 28-day challenge is its informal governance structure. The initiative is organized entirely through Sottiaux’s social media announcements rather than official corporate press releases. Consequently, the precise timeline, governing rules, and qualifying criteria remain fluid. Independent trackers have mapped the operational window from October 5 through November 1, operating on self-determined parameters regarding what constitutes a valid daily shipment.
Furthermore, leadership has not explicitly clarified which subscription tiers will benefit from a potential usage reset, nor have they detailed the precise scope of what constitutes a "full reset" of system quotas. This lack of explicit definition echoes past communication ambiguities. Following the April reset announcement for "all paid plans," numerous Plus subscribers discovered that their standard weekly billing cycles remained unchanged, resulting in confusion regarding whether the special quota relief had been applied to their accounts. OpenAI support later clarified that the global reset did not encompass every distinct quota type across all user profiles.
As the 28-day sprint continues, the engineering team faces the ongoing challenge of balancing aggressive deployment schedules with transparent communication regarding rolling windows, weekly limits, and enterprise-tier allocations. For the developer community, the daily updates represent an unprecedented, high-velocity experiment in software delivery, where the outcomes are measured not only in performance benchmarks, but in the restoration of the computational resources required to sustain modern AI-driven development.
